The infinite Student's t-mixture for robust modeling

Finite mixture models have been widely used for modeling probability distribution of real data sets due to its benefits from analytical tractability. Among the finite mixtures, the finite Student's t-mixture model (SMM) are tolerant to the untypical data (outliers). However, the SMM could not automatically determine the proper number of components, which is important and may has a significant effect on the learned model. In this paper, we propose an infinite Student's t-mixture model (iSMM) to handle this issue. This model is based on the Dirichlet process mixture which assumes the number of components in a mixture is infinite in advance, and determines the appropriate value of this number according to the observed data. Moreover, we derive an efficient variational Bayesian inference algorithm for the proposed model. Through applications in blind signal detection and image segmentation, it is shown that the iSMM possesses the advantages of both the Student's t-distribution and the Dirichlet process mixture, offering a more powerful and robust performance than competing models.     

Object tracking using discriminative sparse appearance model

Object tracking based on sparse representation formulates tracking as searching the candidate with minimal reconstruction error in target template subspace. The key problem lies in modeling the target robustly to vary appearances. The appearance model in most sparsity-based trackers has two main problems. The first is that global structural information and local features are insufficiently combined because the appearance is modeled separately by holistic and local sparse representations. The second problem is that the discriminative information between the target and the background is not fully utilized because the background is rarely considered in modeling. In this study, we develop a robust visual tracking algorithm by modeling the target as a model for discriminative sparse appearance. A discriminative dictionary is trained from the local target patches and the background. The patches display the local features while their position distribution implies the global structure of the target. Thus, the learned dictionary can fully represent the target. The incorporation of the background into dictionary learning also enhances its discriminative capability. Upon modeling the target as a sparse coding histogram based on this learned dictionary, our tracker is embedded into a Bayesian state inference framework to locate a target. We also present a model update scheme in which the update rate is adjusted automatically. In conjunction with the update strategy, the proposed tracker can handle occlusion and alleviate drifting. Comparative results on challenging benchmark image sequences show that the tracking method performs favorably against several state-of-the-art algorithms.     

稀疏贝叶斯字典学习空时机动目标参数估计算法

针对基于稀疏恢复的空时自适应处理(STAP)目标参数估计方法中字典失配导致估计性能下降的问题,该文提出一种基于稀疏贝叶斯字典学习的高精度目标参数估计方法。该方法首先通过目标方位信息补偿多个阵元数据构建联合稀疏恢复数据,然后对补偿后的每个阵元数据利用双线性变换进行加速度和速度项分离。最后构建速度参数和加速度参数的泰勒级数动态字典,对机动目标参数进行高精度贝叶斯字典学习稀疏恢复。仿真实验证明,该方法能有效提高字典失配情况下目标参数估计精度,估计性能优于已有字典固定离散化的稀疏恢复空时目标参数估计方法。     

Self-Adaptive Topic Model： A Solution to the Problem of ＂Rich Topics Get Richer＂

The problem of ＂rich topics get richer＂（RTGR） is popular to the topic models,which will bring the wrong topic distribution if the distributing process has not been intervened.In standard LDA（Latent Dirichlet Allocation） model,each word in all the documents has the same statistical ability.In fact,the words have different impact towards different topics.Under the guidance of this thought,we extend ILDA（Infinite LDA） by considering the bias role of words to divide the topics.We propose a self-adaptive topic model to overcome the RTGR problem specifically.The model proposed in this paper is adapted to three questions：（1） the topic number is changeable with the collection of the documents,which is suitable for the dynamic data;（2） the words have discriminating attributes to topic distribution;（3） a selfadaptive method is used to realize the automatic re-sampling.To verify our model,we design a topic evolution analysis system which can realize the following functions：the topic classification in each cycle,the topic correlation in the adjacent cycles and the strength calculation of the sub topics in the order.The experiment both on NIPS corpus and our self-built news collections showed that the system could meet the given demand,the result was feasible.     

Kernel dictionary learning based discriminant analysis

Sparse representation based classification (SRC) has been successfully applied in many applications. But how to determine appropriate features that can best work with SRC remains an open question. Dictionary learning (DL) has played an import role in the success of sparse representation, while SRC treats the entire training set as a structured dictionary. In addition, as a linear algorithm, SRC cannot handle the data with highly nonlinear distribution. Motivated by these concerns, in this paper, we propose a novel feature learning method (termed kernel dictionary learning based discriminant analysis, KDL-DA). The proposed algorithm aims at learning a projection matrix and a kernel dictionary simultaneously such that in the reduced space the sparse representation of the data can be easily obtained, and the reconstruction residual can be further reduced. Thus, KDL-DA can achieve better performances in the projected space. Extensive experimental results show that our method outperforms many state-of-the-art methods.     

中文文本的LZSS算法实现及研究

文章通过程序设计实现了LZSS压缩算法,并对其进行了改进以适合于中文压缩,改进后的压缩程序测试结构证明改进是有效的,相比于标准LZSS12压缩算法,压缩比有了很大幅度的提高 ,对于中文文本长文件,其最大压缩比已达到20左右,对于英文文本文件的压缩效果也好于LZSS12算法。     

基于优先级扫描Dyna结构的贝叶斯Q学习方法

贝叶斯Q学习方法使用概率分布来描述Q值的不确定性,并结合Q值分布来选择动作,以达到探索与利用的平衡。然而贝叶斯Q学习存在着收敛速度慢且收敛精度低的问题。针对上述问题,提出一种基于优先级扫描Dyna结构的贝叶斯Q学习方法—Dyna-PS-BayesQL。该方法主要分为2部分：在学习部分,对环境的状态迁移函数及奖赏函数建模,并使用贝叶斯Q学习更新动作值函数的参数;在规划部分,基于建立的模型,使用优先级扫描方法和动态规划方法对动作值函数进行规划更新,以提高对历史经验信息的利用,从而提升方法收敛速度及收敛精度。将Dyna-PS-BayesQL应用于链问题和迷宫导航问题,实验结果表明,该方法能较好地平衡探索与利用,且具有较优的收敛速度及收敛精度。     

基于贝叶斯网络工具箱的贝叶斯学习和推理

采用MATLAB语言编制的贝叶斯网络工具箱(Bayesian Networks Toolbox,BNT)可实现贝叶斯网络结构学习、参数学习、推理和构建贝叶斯分类器,此工具箱在贝叶斯学习编程方面非常灵活.介绍了用贝叶斯网络工具箱解决贝叶斯学习和推理问题,并给出了两个实例.     

Context-aware joint dictionary learning for color image demosaicking

Most digital cameras are overlaid with color filter arrays (CFA) on their electronic sensors, and thus only one particular color value would be captured at every pixel location. When producing the output image, one needs to recover the full color image from such incomplete color samples, and this process is known as demosaicking. In this paper, we propose a novel context-constrained demosaicking algorithm via sparse-representation based joint dictionary learning. Given a single mosaicked image with incomplete color samples, we perform color and texture constrained image segmentation and learn a dictionary with different context categories. A joint sparse representation is employed on different image components for predicting the missing color information in the resulting high-resolution image. During the dictionary learning and sparse coding processes, we advocate a locality constraint in our algorithm, which allows us to locate most relevant image data and thus achieve improved demosaicking performance. Experimental results show that the proposed method outperforms several existing or state-of-the-art techniques in terms of both subjective and objective evaluations.     

基于频域信息融合和稀疏贝叶斯学习的高分辨ISAR成像

为了在ISAR成像中更好地压制噪声,消除条纹干扰,提高成像分辨率,本文提出一种基于双向插值处理和频域信息融合的稀疏贝叶斯学习算法,称之为BI?FF SBL算法。该方法首先对回波信号分别进行径向和横向插值预处理,将预处理得到的两份数据通过LA?VB算法进行ISAR成像;然后将得到的两幅ISAR图像通过二维傅里叶变换进入频域,并将两个二维频谱进行信息融合处理,以消除噪声和条纹干扰的相关信息并保留目标结构信息;最后对融合处理后的频谱进行二维傅里叶逆变换,得到最终的ISAR图像。为了验证BI?FF SBL算法的ISAR成像效果,本文进行了基于仿真数据和实测数据的成像实验,并将实验结果与R?D算法、L1?BP算法、LA?VB算法进行对比,发现BI?FF SBL算法在压制噪声和去除条纹干扰方面具有明显的优势,且能提供分辨率更高的ISAR图像。当实验数据信噪比降到0 dB时,BI?FF SBL算法依然能够提供清晰的ISAR图像,明显优于其他三种算法。测试超分辨重构误差的实验结果表明,相比于L1?BP算法和LA?VB算法,BI?FF SBL算法的重构误差更低,在实验数据信噪比为0 dB时,重构信噪比可以达到13.55 dB。     

Unsupervised Bayesian image segmentation using orthogonal series

This paper deals with the problem of unsupervised image segmentation which consists in first mixture identification phase and second a Bayesian decision phase. During the mixture identification phase, the conditional probability density function (pdf) and the a priori class probabilities must be estimated. The most difficult part is the estimation of the number of pixel classes or in other words the estimation of the number of density mixture components. To resolve this problem, we propose here a Stochastic and Nonparametric Expectation-Maximization (SNEM) algorithm. The algorithm finds the most likely number of classes, their associated model parameters and generates a segmentation of the image by classifying the pixels into these classes. The non-parametric aspect comes from the use of the orthogonal series estimator. Experimental results are promising, we have obtained accurate results on a variety of real images.     

基于在线判别式字典学习的鲁棒视觉跟踪

现有子空间跟踪方法较好地解决了目标表观变化和遮挡问题,但是它对复杂背景下目标跟踪的鲁棒性较差。针对此问题,该文首先提出一种基于Fisher准则的在线判别式字典学习模型,利用块坐标下降和替换操作设计了该模型的在线学习算法用于视觉跟踪模板更新。其次,定义候选目标编码系数与目标样本编码系数均值之间的距离为系数误差,提出以候选目标的重构误差与系数误差的组合作为粒子滤波的观测似然跟踪目标。实验结果表明：与现有跟踪方法相比,该文跟踪方法具有较强的鲁棒性和较高的跟踪精度。     

Complex SAR Image Compression Using Entropy-Constrained Dictionary Learning and Universal Trellis Co ded Quantization

In this paper, an Entropy-constrained dic-tionary learning algorithm (ECDLA) is introduced for e?-cient compression of Synthetic aperture radar (SAR) com-plex images. ECDLA RI encodes the Real and imaginary parts of the images using ECDLA and sparse representa-tion, and ECDLA AP encodes the Amplitude and phase parts respectively. When compared with the compression method based on the traditional Dictionary learning al-gorithm (DLA), ECDLA RI improves the Signal-to-noise ratio (SNR) up to 0.66dB and reduces the Mean phase error (MPE) up to 0.0735 than DLA RI. With the same MPE, ECDLA AP outperforms DLA AP by up to 0.87dB in SNR. Furthermore, the proposed method is also suitable for real-time applications.     

Feature discovering for image classification via wavelet-like pattern decomposition

In this paper, we propose a feature discovering method incorporated with a wavelet-like pattern decomposition strategy to address the image classification problem. In each level, we design a discriminative feature discovering dictionary learning (DFDDL) model to exploit the representative visual samples from each class and further decompose the commonality and individuality visual patterns simultaneously. The representative samples reflect the discriminative visual cues per class, which are beneficial for the classification task. Furthermore, the commonality visual elements capture the communal visual patterns across all classes. Meanwhile, the class-specific discriminative information can be collected by the learned individuality visual elements. To further discover the more discriminative feature information from each class, we then integrate the DFDDL into a wavelet-like hierarchical architecture. Due to the designed hierarchical strategy, the discriminative power of feature representation can be promoted. In the experiment, the effectiveness of proposed method is verified on the challenging public datasets.     

基于稀疏稠密结构表示与在线鲁棒字典学习的视觉跟踪

L1跟踪对适度的遮挡具有鲁棒性,但是存在速度慢和易产生模型漂移的不足。为了解决上述两个问题,该文首先提出一种基于稀疏稠密结构的鲁棒表示模型。该模型对目标模板系数和小模板系数分别进行L2范数和L1范数正则化增强了对离群模板的鲁棒性。为了提高目标跟踪速度,基于块坐标优化原理,用岭回归和软阈值操作建立了该模型的快速算法。其次,为降低模型漂移的发生,该文提出一种在线鲁棒的字典学习算法用于模板更新。在粒子滤波框架下,用该表示模型和字典学习算法实现了鲁棒快速的跟踪方法。在多个具有挑战性的图像序列上的实验结果表明：与现有跟踪方法相比,所提跟踪方法具有较优的跟踪性能。     

Educational Data Mining: A review of evaluation process in the e-learning

Due to the growing interest in e-learning as an important process of teaching and learning, new mechanisms to evaluate its pedagogical effectiveness are necessaries. This review describes the scenario of 20 years corresponding to data mining research where the context is the e-learning and the main subject is the evaluation aspect, which is considered a latent problem within this environment. Our goal is to provide an unexplored review of EDM research of the teaching and learning process considering the educational perspective. In order to obtain a more wide and complete review, the search of the bibliographic material was realized with the terms “data mining” and “education” which resulted in 525 articles. As exclusion criterion, articles that did not emphasize the improvement of the teaching and learning process were discarded, resulting in 72 articles. As result of our review, the analyzed papers show that the researches in EDM expanded into several areas and themes, for example, oriented studies on interactions between the educational actors, monitoring, and evaluation of teaching-learning process, administrators’ evaluation about the adopted pedagogical actions, learning risks, and recommendation and recovery of educational media. The review allowed to present perspectives, identify trends and observe potential research directions, such as behavioral research, collaboration, interaction and performance in the development of teaching-learning activities.